Threadripper NUMA and device topology help (VFIO, looking glass)

Hello, I started playing today with my VFIO configuration to extract some more performance theoretically.

AMD Threadripper 2920x
Asrock X399 Taichi
4 x 8GB RAM

First of all, I changed the BIOS from UMA to NUMA as it is supposedly better for latency.

Next, I pinned cores 6 - 11 and their respective threads to my Windows VM and added numatune strict and nodeset 1 in my VM definition.

After running lstopo, I found out that I am using the completely wrong NUMA Nodes. NUMA 0 has my passed-through gpu on, as well as the sata controller that windows runs off (not passed through).

NUMA 1 has the VCPUs, the host GPU, the host NVME and the SR-IOV NIC that both host and guest use.

I am attaching an lstopo photo.

Should I allocate NUMA 0 to the guest? Would allocating core 0 to the guest cause any problems? Or should I try to shuffle devices around the motherboard to try and match each device to its respective NUMA node?

Or does it just not matter and I am making a bigger-than-necessary fuss about it?

Thanks